[PATCH] D66380: [SDAG] Fold umul_lohi with 0 or 1 multiplicand

[Nikita Popov via Phabricator via llvm-commits]

nikic marked an inline comment as done.
nikic added inline comments.


================
Comment at: llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp:4292
 
+  // (umul_lohi N0, 0) -> (0, 0)
+  if (isNullConstant(N->getOperand(1))) {
----------------
RKSimon wrote:
> Do we need a canonicalization of constants to the RHS?
Possibly, but no test case where this would be needed comes to mind. If we start from canonical IR we have constants on the RHS anyway, but even if we start from non-canonical IR, the constant will be moved to the RHS for `ISD::MUL`, before we get here. Possibly it could happen via `UMULO`, though in that case we should probably already canonicalize while it is a `UMULO`.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66380/new/

https://reviews.llvm.org/D66380